Intel Arc A770 16GB - 16GB usable
16GB at 560 GB/s, a lot of capacity for an entry card, and llama.cpp runs on it via Vulkan or SYCL. Suits tinkerers, not people in a hurry.
Specifications
| Brand | Intel |
|---|---|
| Model | Arc A770 16GB |
| Usable VRAM | 16GB |
| Architecture | Xe-HPG |
| CUDA / Stream Processors | 4,096 |
| Memory Bandwidth | 560 GB/s |
| TDP | 225W |
| FP32 TFLOPS | 19.7 |
Current Offers
Used from £400
Prices last updated:
GPUDojo is reader-supported. When you buy through links on our site, we may earn an affiliate commission.
Price History
- eBay£400at high
For AI / LLM Use
Good for 14B models. 30B requires aggressive quantization. Expect a smaller selection of supported local-AI builds than on NVIDIA.
What Models Can It Run?
- 14B Q6_K, 30B Q3_K (tight)
- 14B Q4_K_M, 7B full precision
- 7B Q6_K, 14B Q3_K (tight)
- 7B Q4_K_M only
Estimated Performance
Generation: ~42 tokens/sec
Prefill: ~352 tokens/sec
Recommended Quantisations
- Q4_K_M for 14B models
- Q6_K for 7B-8B models
- Q8 for 7B if VRAM allows
Pros & Cons
Pros
- Consumer card: easy to install, display output
Cons
- 16GB usable VRAM: may need quantization for 30B+ models
- Moderate memory bandwidth: not the fastest for inference
- Smaller local-AI software ecosystem than NVIDIA or AMD
Community Verdict
No community reviews yet for the Arc A770 16GB. Know a good review? Let us know.